Research

This paper examines the impact of climate-induced drinking water scarcity on children's education in rural Ethiopia. I construct a novel, plausibly exogenous measure of well failure by training a random forest classifier on newly collected geo-referenced water point data combined with climatic, hydrological, and geological predictors. The model achieves strong predictive accuracy and yields a panel of predicted well functionality at the enumeration area-month level. Linking these predictions to the Ethiopian Socioeconomic Surveys (2012-2022), I estimate the causal effects of well failure on school absenteeism. I find that well failures increase the likelihood that households report at least one child missing a week or more of school in the last semester by 7 to 8 percentage points, representing a 70 to 80% increase over baseline rates. Effects are present for both boys and girls, with stronger impacts for girls, especially aged 5 to 10. These results are robust to alternative specifications, including different climate controls, and are not explained by health shocks or migration. Instead, well failures trigger a switch from groundwater to surface water leading to an increased water collection burden for children. Finally, I develop a theoretical framework which predicts that when lost hours of schooling exceed the increase in water collection time, child labor intensifies. The empirical evidence suggests that this condition holds: well failures lead to a reorganization of household labor, with a greater reliance on children. 

The number of people without access to water is rising across Sub-Saharan Africa, a challenge further exacerbated by climate change. Frequent droughts highlight the importance of groundwater, often a more reliable water source. As a result, access to wells tapping into groundwater may play a significant role in shaping settlement patterns. This paper investigates whether the presence of wells can act as a pull factor, attracting people to settle nearby. To address this question, I combine information on the location and construction dates of water points across Sub-Saharan Africa with population and nightlight data. It allows me to assess whether well construction leads to increased settlement in their vicinity using a difference-in-differences framework that accounts for staggered adoption.